This does indeed appear to remedy my primary concern with AMP, but I’m still not very comfortable with a vendor-specific standard.
Discourse also doesn’t stand to gain as much from AMP as the average media outlet does, since in our case the amount of “dumbing down” we need to do in order to serve AMP pages means you’re probably only using 5-10% of Discourse’s interactive features. I’d love to serve these types of user segments, but the fact remains that they’re an increasingly smaller minority.
My hope is that AMP eventually turns into a globally recognised standard for low-bandwidth, more secure (because simpler) web content, the same way SPDY paved the way for HTTP/2.
Considering Google rewards speedier pages in general, I’d imagine we stand to gain much more from implementing Service Workers before we revisit AMP.